[ie] Add _search_nextjs_v13_data helper (#13398)

* Fixes FranceTVSiteIE livestream extraction
* Fixes GoPlayIE metadata extraction

Authored by: bashonly

yt_dlp/extractor/common.py

@@ -1783,6 +1783,53 @@ class InfoExtractor:
             r'<script[^>]+id=[\'"]__NEXT_DATA__[\'"][^>]*>', webpage, 'next.js data',
             video_id, end_pattern='</script>', fatal=fatal, default=default, **kw)
 
+    def _search_nextjs_v13_data(self, webpage, video_id, fatal=True):
+        """Parses Next.js app router flight data that was introduced in Next.js v13"""
+        nextjs_data = []
+        if not fatal and not isinstance(webpage, str):
+            return nextjs_data
+
+        def flatten(flight_data):
+            if not isinstance(flight_data, list):
+                return
+            if len(flight_data) == 4 and flight_data[0] == '$':
+                _, name, _, data = flight_data
+                if not isinstance(data, dict):
+                    return
+                children = data.pop('children', None)
+                if data and isinstance(name, str) and name.startswith('$'):
+                    # It is useful hydration JSON data
+                    nextjs_data.append(data)
+                flatten(children)
+                return
+            for f in flight_data:
+                flatten(f)
+
+        flight_text = ''
+        # The pattern for the surrounding JS/tag should be strict as it's a hardcoded string in the next.js source
+        # Ref: https://github.com/vercel/next.js/blob/5a4a08fdc/packages/next/src/server/app-render/use-flight-response.tsx#L189
+        for flight_segment in re.findall(r'<script\b[^>]*>self\.__next_f\.push\((\[.+?\])\)</script>', webpage):
+            segment = self._parse_json(flight_segment, video_id, fatal=fatal, errnote=None if fatal else False)
+            # Some earlier versions of next.js "optimized" away this array structure; this is unsupported
+            # Ref: https://github.com/vercel/next.js/commit/0123a9d5c9a9a77a86f135b7ae30b46ca986d761
+            if not isinstance(segment, list) or len(segment) != 2:
+                self.write_debug(
+                    f'{video_id}: Unsupported next.js flight data structure detected', only_once=True)
+                continue
+            # Only use the relevant payload type (1 == data)
+            # Ref: https://github.com/vercel/next.js/blob/5a4a08fdc/packages/next/src/server/app-render/use-flight-response.tsx#L11-L14
+            payload_type, chunk = segment
+            if payload_type == 1:
+                flight_text += chunk
+
+        for f in flight_text.splitlines():
+            prefix, _, body = f.partition(':')
+            if body.startswith('[') and body.endswith(']') and re.fullmatch(r'[0-9a-f]{1,3}', prefix.lstrip()):
+                # The body isn't necessarily valid JSON, so this should always be non-fatal
+                flatten(self._parse_json(body, video_id, fatal=False, errnote=False))
+
+        return nextjs_data
+
     def _search_nuxt_data(self, webpage, video_id, context_name='__NUXT__', *, fatal=True, traverse=('data', 0)):
         """Parses Nuxt.js metadata. This works as long as the function __NUXT__ invokes is a pure function"""
         rectx = re.escape(context_name)